Prediction of Lignin Content of Manchurian Walnut by BP Neural Network and Near-Infrared Spectroscopy
The lignin as a main component of wood, its content is an important chemical property of wood materials, it has an great effect on the other properties of wood and wood processing and utilization property. In paper making industry, the lignin content is a basis for developing pulp cooking and bleaching process. With the advantages of simple structure, plasticity and obviously superiority in nonlinear data processing, BP neural network and NIR for Manchurian Walnut wood lignin content prediction was investigated in this paper. The original spectra were collected and pretreated with the first derivative. Thriteen typical wave lengths were selected as BP network inputs to establish prediction model for wood lignin content. Model was validated using cross-validation approach. The prediction correlation coefficient (R) is 0.9233 while the root mean square error of prediction (RMSEP) is 0.0179. The results showed that using BP neural network in near-infrared spectroscopy calibration could significantly improve the model performance in order to rapidly and accurately predict wood lignin content.